Real-Time passengers forecasting in congested transit networks considering dynamic service disruptions and passenger count data

Congestion phenomena (e.g., crowding, service interruptions, and atypical demand) increasingly affect complex and interconnected public transport networks, resulting in low levels of services and harming the planned schedules. As a result, public transport operators need a tool to compensate for recurrent and non-recurrent congestion phenomena by recovering the service (e.g., introducing new runs) and notifying passengers about crowding (e.g., through real-time information systems). This study suggests a model that forecasts the volumes of passengers in transit networks, including the effects of events and real-time disruptions. In particular, the model performs a run-based macroscopic transit assignment, computing the elastic route choices of users under the assumption that passengers are fully informed. Moreover, the model corrects its forecasts using real-time count data. The model can also include countermeasures, allowing the operators to test several recovery scenarios on large transit networks faster than in real time.